A pose estimation method of a moving target based on off-board monocular vision

Abstract

Pose estimation plays a vital role in landing guidance of quadcopter unmanned aerial vehicles. In this paper, a continuous-time optimization method is proposed to estimate the pose of a moving object only using an off-board camera. The optimization time that can vary depending upon the desired accuracy and computational speed, and it is finally transferred into solving an ordinary differential equation. Based on the proposed method, a pose estimation system with a camera and a pan-tilt unit is established. The system accuracy and algorithm correctness are tested in an experiment with a moving board as the target. Experimental results indicate that the proposed method can track a moving target and estimate its position and attitude in real time. Therefore, it will be promising to use this method in landing guidance of quadcopter unmanned aerial vehicles.